Amazon has temporarily suspended access to the beta version of the personal assistant "Alexa" and decided to postpone the launch date, which was scheduled for late 2024, to next year, as "Alexa" is still in the development stages.

According to a report from Bloomberg, the main challenge lies in the large language models that have been adopted in the new version, aiming to make "Alexa" better at understanding complex questions. However, it has become less accurate in performing the basic tasks that it previously executed easily, such as setting timers and controlling smart lighting.

The updated version of "Alexa," based on artificial intelligence, was planned to be launched last October, but Amazon decided to delay the date, taking advantage of the event to announce a new production line for "Kindle" devices.

According to previous reports, the upcoming version of "Alexa" is expected to rely on the AI model "Claude" from Anthropic, which will come with a subscription fee.

Sources indicate that the growing interest in "ChatGPT" last year prompted Amazon's CEO, Andy Jassy, to test "Alexa's" competitive capabilities through artificial intelligence. He posed math questions to "Alexa" as if he were a reporter at a sports conference; however, the answers were unsatisfactory, with some even fabricating results of games that did not occur.

Despite this, the personal assistant "Alexa" surpassed the "acceptable quality" standard, leading Jassy and his executive team to decide that their engineers could prepare a beta version by early 2024, but the date has been postponed several times since then.

Some employees have indicated to Bloomberg that the problem is not limited to "Alexa's" performance, but also extends to complexities within Amazon's management structure and a lack of a clear "motivational vision" for the project.